Eyeball Identification and Tracking using Digital Image Processing

Abstract

Real-time eyeball recognition and tracking using digital image processing techniques can give a way of communication. This paper presents the algorithm for tracking the position of eyes in real-time. The following method consists of two steps: face detection and eye-tracking. In vision-based human-computer interaction, skin color for face detection provides a useful cue. Firstly, face detection is done by combining RGB pixel color, HSI (Hue Saturation Intensity) and YCbCr (Luminance Chrome Blue Chrome Red) color based techniques. Then crop the colored face image from the image and divide cropped image into horizontal sections. As eyes are on the upper part of the face so we have used the speeded up robust features (SURF) key points based method on that section and select the strongest point which is located on the eye region. The eye region is segmented by pixel color techniques. Based on the proposed algorithm possibility of false eye detection is reduced. Experimental results show satisfactory performance in a real-time video stream with average accuracy of 97.2%.